项目名称: 高光谱遥感影像光谱解混与超分辨率重建研究及其在地质填图中的应用
项目编号: No.41272359
项目类型: 面上项目
立项/批准年度: 2013
项目学科: 天文学、地球科学
项目作者: 余先川
作者单位: 北京师范大学
项目金额: 90万元
中文摘要: 高准确度高精度的遥感地质填图、矿物填图、地理制图等领域急需高空间分辨率的高光谱遥感影像。超分辨率重建是提高高光谱遥感影像空间分辨率的最根本经济的途径。现有的线性混合像元分解亚像元制图难以正确表征高光谱影像本身的非线性混合问题。为此,本项目根据地物反射机理,充分挖掘影像的光谱相关性,采用非线性独立成分分析、稀疏成分分析等前沿盲源分离技术进行光谱解混,以此来研究高光谱影像的超分辨率重建,核心内容包括:①地物反射机理、地物空间分布相关性、像元与光谱混合模型研究;②基于非线性盲源分离的高光谱影像光谱解混与模型精度评估;③基于先验知识(地物分布空间相关性、高空间分辨率地物成分分析)的超分辨率重建;④高光谱影像的地质填图应用研究。本研究将丰富盲源分离算法理论,推进高光谱影像的实用化,对影像目标精细分类与识别、矿化蚀变信息提取、精准农业病虫害信息提取、农作物估产等领域,具有重大的理论意义与实际应用价值。
中文关键词: 高光谱影像;超分辨率重建;光谱解混;地质填图;非线性独立成分分析
英文摘要: High spatial resolution hyperspectral images are essenial in high-accuracy and high-precision remote sensing geological mapping, mineral mapping, geographic mapping and many other fileds. Super-resolution reconstruction is the most fundamental and economic way to improve the spatial resolution of hyperspectral remote sensing images. However, existing sub pixel mapping based on linear decomposition of mixed pixels is hard to correctly characterize the nonlinear mixed problem in hyperspectral images. Based on the ground-object reflection mechanism, this project aims at exploring the correlation between each image spectral and applying nonlinear blind source separation techniques (e.g. independent component analysis and sparse component analysis) to decompose mixed pixels in the hyperspectral images thus to realize super-resolution reconstruction. The project mainly includes four parts: ①Research of surface reflectance mechanism, spatial distribution correlation of the ground-object and pixel or spectral mixing model; ②Nonlinear blind source separation based spectral decomposition of hyperspectral images and model accuracy assessment; ③Super-resolution reconstruction based on prior knowledge (e.g. spatial distribution correlation of ground-object and component analysis of surface features in high spatial resolution
英文关键词: hyperspectral image;super-resolution reconstruction;spectral unmixing;geogical mapping;nonlinear ICA